The present invention relates in general to digital processing and more specifically to networked computing environments and a model for amortization of hosting cost and for fine-tuning Quality of Service (QoS) in networked computing environments that are hosted in centralized locations.
Today's computing environments are almost always networked and in particular are connected to the Internet. Logistical reasons such as space management, physical security, power distribution, and communication access require computing environments of any reasonable size to be located in Network Operation Centers (NOCs).
Also known as Data Centers or Hosting Centers, NOCs act as the physical hubs of the Internet, where computing systems are located in a large scale and have access to communication bandwidth. An NOC may directly serve a single organization such as a large corporation, or serve multiple organizations, both large and small, through an intermediary organization responsible for the maintenance and administration of the NOC. In either case, an NOC is physically divided into multiple isolated units of physical space called ‘cages’, where each cage is self-contained in terms of access to power lines and communication lines. A cage can hold a multitude of ‘racks’, each of which is a shelf, usually satisfying an industry-standard specification for size.
A rack is a rectangular frame with four vertical angles (one at each corner) and one or more matching pairs of supporting braces on opposite sides. Each rack typically has 42 units of space and each unit is usually referred to as ‘1U’. Thus, at most, 42 computers of size 1U can be stacked up in a single rack. Other possible arrangements of computers on a rack may be, for instance, a rack that can hold at most 21 computers of size 2U, a rack that can hold at most 10 computers of size 4U, and so on. These computers are specially packaged to fit on a rack and are referred to as ‘rack-mountable’ boxes. The height (or thickness) of a box has been standardized in the industry to be a multiple of 1U.
A major component of the cost associated with hosting in a data center is the cost of the space used. One limitation of the prior art physical arrangement of computers on racks is that only boxes of certain sizes can be hosted, i.e., the quality of the computing service is tightly coupled to the physical sizes of rack-mountable boxes. Thus, NOCs are restricted in the quality of computing offered in the racks.
FIG. 1 illustrates a prior art arrangement of computers on a rack 12 within an NOC 10. Typically, forty computers 14 of 1U size are mounted on a rack so that each computer occupies 1U of the height of the rack and the top most two 1U units are used for power extension modules. Power cables 22 are connected to the back of computers 14 and extend to a power extension module 30 on the top of rack 12. In prior art systems, access to both the front and the rear of computers 14 is required. One side (typically the rear side) includes the power controls for administration (power-on, power-off, and reboot) and the other side (typically the front side) includes the floppy drives or CD drives for software installations/upgrades (in particular, installation of OS or booting). Both sides need to be accessed in prior art systems because activities associated with computers 14, such as administration and installations, are done manually making the accessibility of both sides of computers 14 necessary.
Typically, NOCs receive customer requests for a network with a certain amount of computing power. The NOC then manually sets up the network with computers 14. A multiple of 1U of computing power is allocated to a customer. Smaller increments of computing power are not available because the entire 1U computer is allocated. Thus, the allocation of computing power is not tunable in increments of computing power under 1U.